Added after our earlier coverage. The original source date is shown below; this is not a new announcement today.

What happened

Microsoft lists all three through Foundry. Transcription supports 60 languages and costs an introductory $0.54 per audio hour through year-end. Both speech-generation models support 23 languages and 26 locales, priced at $22 and $15 per million characters respectively. The announcement offers hosted services, not downloadable weights.

iNC / Analysis

Why it matters

Live transcription and speech generation can shorten a voice agent’s response loop. The different language counts matter: recognizing a language does not mean the paired voice model can answer in it. Vendor latency figures need verification in complete applications.

What to watch

Watch language-by-language accuracy, end-to-end delay and pricing after the introductory period.

Sources & verification

Source links are not a count of independent confirmations.

01
Microsoft AI — streaming and voice releasePrimary source · microsoft.ai · Oct 1, 2026
Report a correction

Help us get it right

Include the disputed statement and, if possible, a supporting source URL. Please do not send personal or sensitive information.